Race/Ethnicity and Reference Equations for Spirometry

To the Editor:

We read with interest two papers in the Journal arguing that lung-function prediction equations should be neutral with regard to race and ethnicity. We agree that race is a socio-political construct and we must eliminate racial biases in health care, but we disagree with the approach of McCormack and colleagues (1) and of Elmaleh-Sachs and colleagues (2) who redefine normal values for spirometry to address this issue.

By ignoring a subject’s ancestry when evaluating lung function, McCormack and colleagues (1) found spirometry better correlated with subsequent overall mortality in National Health and Nutrition Examination Survey (NHANES) III data. In NHANES III, those of African ancestry were on average younger but had an age-adjusted mortality that was worse than that for people of European ancestry, which fits with Centers for Disease Control and Prevention data (3). In the United States, the small proportion of total deaths related to chronic lower respiratory disease differ significantly between people of African versus European ancestry (3.3% vs. 6.4%, respectively) (4), but the authors attempt to account for all of the difference in all-cause mortality by manipulating lung function data. Using global prediction equations (Global Lung Function Initiative “other” [5]) that combine all ancestries makes the lung function for those of African ancestry appear worse and for those of European ancestry appear better. The resultant improvement in the correlation of FEV₁ z-score with overall mortality is used to justify using global prediction equations. However, the great majority of overall mortality differences are not related to lung function. The different disease spectrum and limitations in both the access to and the delivery of health care for people of African ancestry, that are not accounted for by socio-economic adjustments (6), are not addressed by the authors. Because there are different numbers in the two groups being compared in the study, the probability distribution graphs in the study should use percentage of people rather than numbers of people, as demonstrated in Figure 1, which shows both groups had z-scores approximately centered around unity, which is to be expected from a general population cohort compared with a healthy reference population. Using geographic ancestry-specific equations (5) does not produce a bias between the two groups with the distributions of initial z-scores for FEV₁ in the two groups being remarkably similar.

Figure 1.

Data replotted from McCormack et al. (1) as percentage of the group rather than the number of people for African ancestry (blue lines) and European ancestry (orange lines).

However, using globally based z-scores that combine all geographic ancestries skews the two groups in different directions. The European ancestry group are shifted to higher and the African ancestry group are shifted to lower FEV₁ z-scores, making the general population of African ancestry cohort appear to have FEV₁ lower than the reference African ancestry cohort (suiting the authors’ thesis), while the general population of European ancestry cohort then appear to have FEV₁ higher than the reference European ancestry cohort, which is highly improbable. As the authors note, using global reference equations for spirometry is potentially prejudicial to patient care for both people of African ancestry and European ancestry with a risk of overdiagnosis of respiratory disease in the former and underdiagnosis in the latter.

Elmaleh-Sachs and colleagues (2) also looked at survival- and event-related data in NHANES III data to make a conclusion that race–neutral lung-function prediction equations are the best way forward. Their analysis not only suffers from the problems outlined above, but they also used percentage of predicted lung function values in their analysis. This is a flawed methodology that is not supported by the American Thoracic Society or European Respiratory Society in making assessments about lung function (7). It retains sex, age and size bias and assumes a proportionality in severity which is not proven. Because it retains a size bias, it will include a geographic ancestry bias. Percentage of predicted also ignores the degree of scatter found in normal subjects which varies with sex and geographic ancestry (8).

Improving the chance that people of African-American ancestry will receive equitable health care is unlikely to be achieved by reducing the precision of spirometry reference values. It is important to distinguish between genuine racism in healthcare and the effects of geographic ancestry on lung function. Anthropomorphic differences in sitting height to standing height account for at least 35% of the discrepancy in lung function between African Americans and White subjects with a further 2.5% to 7.5% relating to poverty and 2.0% to 4.7% to education (9). A better account and understanding of these substantial anthropomorphic differences is needed. We believe that the above authors’ conclusions to ignore the differences in the relation between lung function and the sex, age, and standing height for people of different geographic ancestry is not justified from their findings. Neither paper has shown an improvement for individual patients, and their approach could lead to racism from prejudicial judgments being made about whether an individual's lung function is within the range expected for someone of the same geographic ancestry. This will obscure the true causes for the worse overall mortality for people of African ancestry, which must be addressed so that worse prejudicial outcomes do not continue.